Ablate just added

cais/wmdp-corpora

published by @cais

Refusal-direction material, activation probes, steering and concept-erasure data.

task_categories:text-generationlanguage:enlicense:mitsize_categories:10K<n<100Kformat:parquetmodality:textlibrary:datasetslibrary:dask +4 more
Lifetime downloads
65.4K
Last 30 days
1.8K
Likes
6
Size
921.0 MB

Description

Dataset Card for WMDP Corpora

The Weapons of Mass Destruction Proxy (WMDP) Corpora includes all of the corpora used to perform unlearning on WMDP-Bio and WMDP-Cyber.

See our paper, website, and GitHub for more details!

The corpora are also available at the following mirrors with password wmdpcorpora: 1, 2

The bio forget corpus must be requested separately; please visit this form.

cyber-retain-corpus and cyber-forget-corpus

The forget and retain corpora… See the full description on the dataset page: https://huggingface.co/datasets/cais/wmdp-corpora.

Used by abliterated models none yet

No abliterated models in our catalog have declared this dataset in their YAML metadata yet. This can mean the dataset is used but not declared, is used outside abliteration workflows, or is new. The link will populate automatically as models are indexed.

Metadata

License
mit
Language
en
Created on HF
2024-03-05 · 2 years ago
Size category
10K<n<100K
Modalities
text
Formats
parquet
Libraries
datasetsdaskmlcroissantpolars

Related datasets same stage · Ablate

Read further

Total size
921.0 MB
Files
9
Formats
parquet, png
Head
daf89fa
Last updated on HF
2024-04-25

Download the dataset

Copy any of these snippets into your notebook or terminal. All three fetch directly from Hugging Face using your own credentials.

Recommended: datasets library (Python)
from datasets import load_dataset
ds = load_dataset("cais/wmdp-corpora")
Raw snapshot (Python)
from huggingface_hub import snapshot_download
snapshot_download(repo_id="cais/wmdp-corpora", repo_type="dataset")
git clone (requires git-lfs)
git clone https://huggingface.co/datasets/cais/wmdp-corpora

Files by folder from our last crawl

top level 3 files 293.2 KB
.gitattributes 2.3 KB file download
dataset.png 287.1 KB png download
README.md 3.8 KB md download
bio-retain-corpus/ 4 files 883.6 MB
train-00000-of-00004.parquet 221.7 MB parquet download
train-00001-of-00004.parquet 219.6 MB parquet download
train-00002-of-00004.parquet 221.6 MB parquet download
train-00003-of-00004.parquet 220.7 MB parquet download
cyber-forget-corpus/ 1 file 10.8 MB
train-00000-of-00001.parquet 10.8 MB parquet download
cyber-retain-corpus/ 1 file 26.4 MB
train-00000-of-00001.parquet 26.4 MB parquet download

No README stored for this dataset yet.

Download from HuggingFace

cais/wmdp-corpora

Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration